We have seen that the changes undergone by ‘the same language’ over a given period of time can be extracted from a comparison of its successive states, the states chosen for this purpose being in practice separated from one another by gaps of at least several hundred years. The past states of the language subjected to comparative analysis may be either attested in written documents or the products of reconstruction based on the comparison of related languages or dialects. We have also seen that the diachronic concepts of ‘the same language’ and of ‘related languages’ are capable of formal definition. In the same way, the ‘change’ undergone by a language may be defined as the sum total of the unidirectional rules which must be set up in order to account for the differences between its successive states, or between a reconstructed ancestor and each of its descendants. We have, however, so far said nothing about the actual mechanism of language change, and in order to deal with this aspect we must now turn from the examination of grammars to the behaviour of individual speakers. We have chosen this order deliberately, for unless we are first familiar with the formal properties of diachronic rules we will be in no position to recognize and interpret the phenomena of change as it actually takes place, that is to say as it is reflected in the behaviour of the individual members of a speech community.